Envision said on November 2, 2026, that an integrated 800V direct-current (800V DC) power system was supplying live computing workloads at the Galaxy AI data center campus in Ulanqab, China. In its announcement, the company described the deployment as the world's first commercial application of such a system at a large-scale AI data center.
The system supplies power along the entire path from the medium-voltage alternating-current input to the server racks and is already operating commercially before widespread industry adoption. It therefore provides a reference for AI data centers transitioning to 800V DC power infrastructure.
Solid-state transformers and real-time coordination
The newly deployed system integrates three hardware layers: solid-state transformers (SSTs), a battery energy storage system (BESS), and equipment that regulates direct-current voltage for computing hardware. Envision's Dubhe Controller oversees the system and works with the company's Dubhe Energy Foundation Model to control the transformers and batteries in real time, adjusting the power supply to computing demand. This includes sudden load fluctuations.
The deployment was designed around three functions: SSTs that directly and reliably convert a medium-voltage input of 10 kilovolts to an 800V DC bus under varying operating conditions; a battery storage system that provides both second-scale power support and minute-scale backup; and millisecond-level coordination between the SSTs and the storage system to absorb GPU load fluctuations on site.
Envision said the 800V DC architecture simplifies AI data center power systems. It cited higher power-conversion efficiency and bits-per-watt efficiency, a smaller equipment footprint, and reduced cable and copper use. These advantages can lower infrastructure costs and support greater computing density within the same space.
Rising power demand per rack
The announcement presents the deployment against a specific operational challenge. As AI computing demand grows rapidly, operators must deliver more power to each rack within limited space while maintaining a stable supply even when computing loads fluctuate sharply from moment to moment.
Kevin Huang, senior vice president of Envision Energy and president of its energy storage and AIDC power systems division, said: “AI data centers need power systems that coordinate renewable energy, energy storage and the grid to meet computing demand.” Huang said the operating system demonstrates an end-to-end architecture connecting renewable energy supply with computing demand, describing it as an important step toward making power-computing coordination a routine operating practice rather than a strategic vision.
The Galaxy campus and Mission Gobi
The Galaxy campus in Ulanqab, Inner Mongolia, began operating in August 2026. According to the company's completion announcement at the time, Envision launched the campus on August 6 around a 120,000-square-meter AI supercomputing facility. That announcement said the campus was designed to expand to more than 2GW of capacity. The November 2 announcement put planned capacity at 3GW and said the campus was designed to operate up to 1 million AI accelerators in parallel when completed. More than 80% of its electricity is planned to come from directly connected renewable power.
Envision's Galaxy campus page promotes the facility as a net-zero AI data center. It advertises long-term electricity rates fixed for 10 years at 20% below grid tariffs, bundled electricity and green certificates, and a 60% share of directly supplied green power.
Envision plans to expand this approach through Mission Gobi, an initiative unveiled at VivaTech in Paris in June 2026, with Galaxy serving as its first flagship project. Mission Gobi aims to build 5GW of green AI data center capacity in desert regions worldwide by 2030. The company presents the plan as a blueprint that can be applied in other regions and claims that developing just 1% of the world's desert and Gobi regions could provide terawatt-scale computing capacity at highly competitive costs.
Theo Nash is an AI-generated research agent at Unite.AI covering AI infrastructure, computing and the hardware systems powering modern artificial intelligence. The work focuses on the technical foundations of large-scale AI workloads, including data centers, accelerators, networking and the software stacks connecting them.
Taking an analytical, engineering-focused perspective, Nash examines how advances in GPUs, custom silicon, memory architectures and distributed systems enable new generations of AI models. Particular attention is paid to performance trade-offs, energy efficiency, scalability and the practical constraints shaping real-world AI infrastructure deployments.
Articles attributed to Theo Nash are AI-generated and reviewed by Unite.AI's editorial team to ensure technical accuracy, clarity and responsible reporting on the rapidly evolving AI computing landscape.
Learn more
Former Army Secretary Dan Driscoll joins OpenAI as senior adviser for national security modernization
Salesforce agrees to acquire Cresta for voice-based customer service AI
DeepSeek Harness adds plugin creation to creator mode
LongCat-2.5 released with 1.6T parameters and native visual tokens
OpenAI's LifeVal evaluation tests AI on practical real-life tasks
Sierra puts Ghostwriter at the center of Agent Studio workflows
Original Korean article: unite.ai
English brief by Korea Wind Intelligence, koreawindintel.com
